Jensen Huang, NVIDIA CEO, told folks durinng a Q&A session at this year’s GTC that he estimated a roughly 20% increase in performance with gate-all-around (GAA) transistors, and that this will play an important part in the advancement of the Feynman generations, which is expected to roll out by 2028.
Semiconductor makers still use the FinFET transistor design, which has been employed in nodes as recently as 3nm. GAA transistors, in this case, is used for transistor gates that have shrunk beyond 3nm. Intel’s upcoming 18A is one such example, with its upcoming Panther Lake lineup set to be built using its own RibbonFET process.
As for NVIDIA, it’s upcoming GPU generation, Vera Rubin, will still be built using TSMC’s N3 node, with products from the architecture only set to be shipped out to enterprise clients by next year. With that in mind, we think it’s safe to say that we’re not expecting to hear anything regarding Feynman or the GPU brand using GAA transistors at least until 2027.
(Source: Techspot)
